Increasing emphasis by insurers on cost-containment and appropriateimaging utilization has prompted new interest in expert systemsthat guide medical decision-making. One of these, Phoenix RadiologyConsultant, is positioned to help referring physicians
Increasing emphasis by insurers on cost-containment and appropriateimaging utilization has prompted new interest in expert systemsthat guide medical decision-making. One of these, Phoenix RadiologyConsultant, is positioned to help referring physicians selectand request imaging procedures.
Phoenix was developed by Dr. Charles Kahn, an associate professorof radiology and medical informatics at the Medical College ofWisconsin, Milwaukee. The program is the only one of its kindto undergo clinical trials.
The program, in use for two years at the University of Chicago,has been favorably received by residents. Currently undergoingclinical revision by Kahn and colleagues at Yale University, Phoenixwill be available for use within six months, Kahn said.
Phoenix is a rule-based, if-then system. It uses a graphicpresentation format to present radiological workup strategiesfor 54 common clinical problems. The information is based on radiologicaland clinical texts, which are referenced in the program.
The data can also be tailored to permit maximum guidance dependingupon the imaging resources available in a given setting, Kahnsaid.
Although it is currently positioned as an educational toolfor residents and attending physicians, the goal is to ultimatelyintegrate Phoenix into existing hospital and radiology informationsystems, Kahn said.
"Eventually, we'd like to have referring physicians requestingthe radiology exam through their information systems, with Phoenixacting as surrogate radiology expert that guides them toward thecorrect exam," he said.
The program would work in concert with radiology informationsystems that integrate scheduling, reporting and patient records.Before the product reaches that stage, however, Kahn will be restructuringPhoenix from a rule-based system to one that is case-based.
While appropriate for educational purposes, rule-based systemshave inherent limitations for clinical applications. Case-basedreasoning attempts to model the way humans learn from experience,he said.
"We learn how to do things today because we remember howwe did things yesterday," he explained. "If hittinga nail with a pair of pliers doesn't work but hitting it witha hammer does, the next time you will use a hammer first. Case-basedreasoning systems similarly learn from experience. It's a verypowerful technology."
The tricky part of attracting users to such a program liesin conveying the expert information. Most physicians do not wanta computer to tell them what to do, Kahn said, and the Phoenixprogram reflects that. The program analyzes the information enteredand critiques it in comparison to existing alternatives.
For example, a referring physician requests an ultrasound scanwhen what he or she really needs is a CT scan.
"If you go ahead with the ultrasound, and then subsequentlydo a CT, you have to hope the insurance company will pay for theunnecessary ultrasound," he said. "It would be betterto be able to say when the ultrasound is requested that a CT mightbe better, and why, with referenced material to back it up. It'sprospective quality control."
In an era of utilization review, expert systems that guideappropriate use appear ripe for development, Kahn said. One ofthe ways imaging centers and radiology departments can move forwardis by promoting knowledge about imaging procedures. The otheris by viewing radiology consultation as an integrated processthat goes beyond exam performance and image interpretation.
"In radiology we tend to concentrate on performance andinterpretation instead of procedure selection and report communication,"he said. "But more attention is being focused on those areas."
Can AI Predict Future Lung Cancer Risk from a Single CT Scan?
May 19th 2025In never-smokers, deep learning assessment of single baseline low-dose computed tomography (CT) scans demonstrated a 79 percent AUC for predicting lung cancer up to six years later, according to new research presented today at the American Thoracic Society (ATS) 2025 International Conference.
What if Radiology Turns Out Exactly the Way We Predict it Will?
May 19th 2025Whether it is reimbursement cuts or continued attempts to push non-radiologist image interpretation, where do we draw the line between inspired protest and misspent energy criticizing things that are doomed to fail or things we have no control over?
What a New PSMA PET/CT Study Reveals About Local PCa Treatment and High-Risk Recurrence
May 16th 2025For patients at high-risk for biochemical recurrence of prostate cancer, PSMA PET/CT findings revealed that 77 percent had one or more prostate lesions after undergoing local radiation therapy or radical prostatectomy, according to a recent study.